train multiprocess issues

[AirBall0]

Hi, thanks for your excellent work and I am interest in. But when I try to run your code(don't change anything), the first iteration is ok, and it can print the information about reward and so on. when enter into 2 iter, the main process is also ok, but the followed childs process is always run without cpu usage, and finally the problem was located at policy_vs_net, It's gonna stay stuck here without forward output, so have you meet this problem? Thanks!